Application of wearable biosensors to construction sites. II: Assessing workers' physical demand

The construction industry is one of the world's most labor-intensive industries. In it, workers are challenged almost every day by highly demanding physical tasks. Although current methods [e.g., the National Institute for Occupational Safety and Health (NIOSH)] to investigate the physical demands of various tasks provide valuable information with which to evaluate certain manual handling tasks, they may be limited to consideration of unique characteristics of each individual (e.g., physiological characteristics) and environmental conditions (e.g., ambient temperature and humidity). In other words, given the same task, different workers experience different levels of exertion. To address this problem, the objective of this research is to develop a procedure for automatic predictions of demand levels based on physiological signals collected from workers. To achieve the objective, workers' physiological signals were captured using a wristband-type biosensor while they performed regular tasks in the field. Various physiological responses were extracted from the artifact-corrected physiological signals. The rate of energy expenditure, estimated using an energy-expenditure prediction program (EEPP), was used as a baseline to separate tasks into low-, moderate-, and high-intensity activities. Then, a supervised-machine-learning model was trained by applying a Gaussian kernel support vector machine. The results led to a prediction accuracy of 90% in recognizing low and high physical-intensity levels and 87% for low, moderate, and high physical-intensity levels. The main contribution to the body of knowledge is the development of an automatic and noninvasive method for assessing workers' physical demands in the field. This study will contribute to improving construction workers' productivity, safety, and general well-being through the early detection of highly physically demanding tasks in the field.